Hot to test for _FORTIFY_SOURCE with Autoconf

Hot to test for _FORTIFY_SOURCE with Autoconf

[Jeffrey Walton]

Hi Everyone,

I'm not sure if this is a GCC or Autoconf question. Please point me in
the right direction.

I'm trying to test for availability of _FORTIFY_SOURCE=2. Setting a
CPPFLAG is not enough since the compiler will happily pass it down.

Does the compiler make the information available anywhere?

Here's a small Autoconf test, but I'm not sure how to craft the program.

  OLD_CPPFLAGS="$CPPFLAGS"
  CPPFLAGS="$CPPFLAGS -D_FORTIFY_SOURCE=2"

  AC_MSG_CHECKING([for -D_FORTIFY_SOURCE=2])
  AC_LINK_IFELSE(
    [AC_LANG_SOURCE([int main(void) {return 0;}])],
    [AC_MSG_RESULT([yes]); MY_CPPFLAGS="$MY_CPPFLAGS -D_FORTIFY_SOURCE=2"],
    [AC_MSG_RESULT([no])]
  )

How do I check for _FORTIFY_SOURCE=2?

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Jonathan Wakely]

On Tue, 17 Mar 2020 at 22:15, Jeffrey Walton via Gcc-help
<gcc-help@gcc.gnu.org> wrote:
>
> I'm trying to test for availability of _FORTIFY_SOURCE=2.

What does that mean?

If you mean you're trying to test whether setting that has any effect,
that's determined by Glibc not GCC.

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Jeffrey Walton]

On Tue, Mar 17, 2020 at 8:11 PM Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
>
> On Tue, 17 Mar 2020 at 22:15, Jeffrey Walton via Gcc-help
> <gcc-help@gcc.gnu.org> wrote:
> >
> > I'm trying to test for availability of _FORTIFY_SOURCE=2.
>
> What does that mean?

I'm trying to determine if the compiler supports FORTIFY_SOURCE.

When _FORTIFY_SOURCE=2 is used, *_chk functions are used when the
compiler can determine the destination buffer size. Effectively the
compiler inserts those Microsoft safer functions that the libc folks
rejected.

I think it is related to object size checking.

Jeff

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Jonathan Wakely]

On Wed, 18 Mar 2020 at 00:16, Jeffrey Walton <noloader@gmail.com> wrote:
>
> On Tue, Mar 17, 2020 at 8:11 PM Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
> >
> > On Tue, 17 Mar 2020 at 22:15, Jeffrey Walton via Gcc-help
> > <gcc-help@gcc.gnu.org> wrote:
> > >
> > > I'm trying to test for availability of _FORTIFY_SOURCE=2.
> >
> > What does that mean?
>
> I'm trying to determine if the compiler supports FORTIFY_SOURCE.
>
> When _FORTIFY_SOURCE=2 is used, *_chk functions are used when the
> compiler can determine the destination buffer size. Effectively the
> compiler inserts those Microsoft safer functions that the libc folks
> rejected.

I don't think the compiler inserts anything. _FORTIFY_SOURCE enables
features in Glibc, not GCC.

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Martin Sebor]

On 3/17/20 6:16 PM, Jeffrey Walton via Gcc-help wrote:
> On Tue, Mar 17, 2020 at 8:11 PM Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
>>
>> On Tue, 17 Mar 2020 at 22:15, Jeffrey Walton via Gcc-help
>> <gcc-help@gcc.gnu.org> wrote:
>>>
>>> I'm trying to test for availability of _FORTIFY_SOURCE=2.
>>
>> What does that mean?
> 
> I'm trying to determine if the compiler supports FORTIFY_SOURCE.
> 
> When _FORTIFY_SOURCE=2 is used, *_chk functions are used when the
> compiler can determine the destination buffer size. Effectively the
> compiler inserts those Microsoft safer functions that the libc folks
> rejected.
> 
> I think it is related to object size checking.

In GCC 10 you can test __has_builtin (__builtin_object_size).  In GCC
versions that don't support __has_builtin using __builtin_object_size
will either compile or not, depending on whether it's supported.  It
doesn't depend on any macros.

Martin

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Jim Wilson]

On Tue, Mar 17, 2020 at 5:16 PM Jeffrey Walton via Gcc-help
<gcc-help@gcc.gnu.org> wrote:
> When _FORTIFY_SOURCE=2 is used, *_chk functions are used when the
> compiler can determine the destination buffer size. Effectively the
> compiler inserts those Microsoft safer functions that the libc folks
> rejected.

Then that is how you test for it in a configure script.  Just write a
trivial program that will generate a call to a *_chk function if
_FORTIFY_SOURCE=2 is supported, compile it, and then check for that,
either by greping the assembly source or running nm on the object
file.

Jim

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Jim Wilson]

On Tue, Mar 17, 2020 at 6:49 PM Martin Sebor via Gcc-help
<gcc-help@gcc.gnu.org> wrote:
> > On Tue, Mar 17, 2020 at 8:11 PM Jonathan Wakely <jwakely.gcc@gmail.com> wrote:
> > I think it is related to object size checking.
>
> In GCC 10 you can test __has_builtin (__builtin_object_size).  In GCC
> versions that don't support __has_builtin using __builtin_object_size
> will either compile or not, depending on whether it's supported.  It
> doesn't depend on any macros.

That is a good way to test if the compiler has object size checking
support, but it is no guarantee that the C library has FORTIFY_SOURCE
support.  If it is fortify source support that you need, then you need
to explicitly check for that.  If it is just the object size checking
support you need, then you can do the check this way.

Jim

Re: Hot to test for _FORTIFY_SOURCE with Autoconf

[Segher Boessenkool]

On Wed, Mar 18, 2020 at 03:17:27PM -0700, Jim Wilson wrote:
> On Tue, Mar 17, 2020 at 5:16 PM Jeffrey Walton via Gcc-help
> <gcc-help@gcc.gnu.org> wrote:
> > When _FORTIFY_SOURCE=2 is used, *_chk functions are used when the
> > compiler can determine the destination buffer size. Effectively the
> > compiler inserts those Microsoft safer functions that the libc folks
> > rejected.
> 
> Then that is how you test for it in a configure script.  Just write a
> trivial program that will generate a call to a *_chk function if
> _FORTIFY_SOURCE=2 is supported, compile it, and then check for that,
> either by greping the assembly source or running nm on the object
> file.

glibc itself uses

#if defined _FORTIFY_SOURCE && _FORTIFY_SOURCE > 1


Segher